Extensive pancreatic resection for small-duct chronic pancreatitis is often required for pain relief, but the risk of diabetes is a major deterrent. Incidence of pain relief, prevention of diabetes, and identification of factors predictive of success were the goals in this series of 48 patients who underwent pancreatectomy and islet autotransplantation for chronic pancreatitis. Of the 48 patients, 43 underwent total or near-total (> 95%) pancreatectomy and 5 underwent partial pancreatectomy. The resected pancreas was dispersed by either old (n = 26) or new (n = 22) methods of collagenase digestion. Islets were injected into the portal vein of 46 of the 48 patients and under the kidney capsule in the remaining 2. Postoperative morbidity, mortality, pain relief, and need for exogenous insulin were determined, and actuarial probability of postoperative insulin independence was calculated based on several variables. One perioperative death occurred. Surgical complications occurred in 12 of the 48 patients (25%): of these, 3 had a total (n = 27); 8, a near-total (n = 16); and 1, a partial pancreatectomy (p = 0.02). Most of the 48 patients had a transient increase in portal venous pressure after islet infusion, but no serious sequelae developed. More than 80% of patients experienced significant pain relief after pancreatectomy. Of the 39 patients who underwent total or near-total pancreatectomy, 20 (51%) were initially insulin independent. Between 2 and 10 years after transplantation, 34% were insulin independent, with no grafts failing after 2 years. The main predictor of insulin independence was the number of islets transplanted (of 14 patients who received > 300,000 islets, 74% were insulin independent at > 2 years after transplantation). In turn, the number of islets recovered correlated with the degree of fibrosis (r = -0.52, p = 0.006) and the dispersion method (p = 0.005). Pancreatectomy can relieve intractable pain caused by chronic pancreatitis. Islet autotransplantation is safe and can prevent long-term diabetes in more than 33% of patients and should be an adjunct to any pancreatic resection. A given patient's probability of success can be predicted by the morphologic features of the pancreas.